Visual perception evolved in a world of objects many of which are bounded by smooth closed contours. Especially with the three-dimensional dataset that need to be marked the locations of filaments to get the region of interest. There also exist other structures with high values (noise) or we may loose some information within the image through contour detection process. For instance, a simple global thresholding method based on gradient edge detection will give us poor results even when if we choose an optimal threshold value. Indeed, there exist many other structures with high grey level values. Global thresholding may keep noise if threshold parameter is too low, and we may loose some image information if threshold parameter is too large. The paper describes a new contour closing method, a local operator based on the so-called Chamfer distance transformation. Inside we give a quick overview of the original philosophy and motivation of this algorithm. In addition, some numerical simulations will be presented and applied to the 3D biological mice vessel network.
